The new AMD roadmap restores a bit of confidence
Analysis Still a long way to go

By Charlie Demerjian: Wednesday, 07 May 2008, 8:14 PM
http://www.theinquirer.net/gb/inquirer/news/2008/05/07/amd-roa...stores-bit

And yes AMD is planning on a 6-core, 8-core and 12-core - "NATIVE" for late 2008, into 2009 & 2010. Interesting they are sticking with the Native which IMO is a good idea.

Super XP said:
That word had a lot to do with why AMD is where they are now (stubbornness, elitist attitude). If they keep pushing it, they might as well declare bankruptcy now.


Basically, that article is saying AMD's cores suck and they only way they can fix it is to increase the core count. Personally, I think it's a bad idea because manufacturer costs more, it consumes more power, produces more heat, and ultimately shrinks how much money AMD could make per chip which is already very low. AMD needs to forget the enthusiast/server markets and focus on budget/mainstream for the time being.

Once Nehalem hits the shelf, AMD simply won't be able to afford to be in that market anyway. They are getting outspent by billions of dollars. They need to focus on the sectors of the market that they are competitive in. Eventually they'll be prepared to take on Intel again but for the time being, that's a bad idea.


Edit: Simply put, I don't even look at Intel for the ultra cheap computers. A Brisbane spanks a Celeron any way you shake it and is much cheaper than the Core-based processors. They are very competitive for the budget computer right now and they should really, really push that.

There is nothing wrong with AMD’s approach in creating/designing the so called “Native” CPU’s. Their current architecture permits them to do so, but what really sucks is AMD’s management and unfortunately s**t falls downwards and hits the little guy at the bottom that really is not responsible for this AMD fiasco. The CEO needs to find another job because he already failed AMD.

You do have valid points, AMD needs to stop chasing Intel for performance leader and start concentrating on what they do best until they can dig themselves out of the dirt. Though buying out ATI was an extremely smart move. They are generating a lot of money out of ATI due to the fact its being run a lot better than AMD is.

Super XP said: They're thriving on that word and when the product comes out that pushes that word so hard flops harder than a blue whale, the result is very little consumer trust. They need to stop using that word and they need to do it now. They're only digging their hole deeper.


Super XP said: a) GPUs bring in far less revenue than CPUs do so no, they are still a long ways from getting their initial investment costs back.
b) By buying out ATI, they forced Intel to dig up the Timna concept (system-on-a-chip). That would have been a great move at the height of Athlon 64 but, their timing couldn't have been any worse. Smart move: yes. Right time, no. They effectively pushed the market in a direction they aren't even close to being ready to compete in. Now they are required to spend money on something no one else wanted to do when they don't have money in the first place to do so.
c) Had NVIDIA not fallen into another slump in 2007/2008, ATI could have been hurting as much, if not more so than AMD.


Edit: Fixed first quote.

the hell did you come from Super? lol

Great points FORD. You really should consider writing a letter to AMD's CEO despite his uselessness as of late.
Sean Costello said:
I came from outer space
My 9 month old keeps me very busy. Darn baby never likes to sleep lol

Super XP said:

well that explains a lot...

Super XP said:


Oh ok lol, congrats on the baby, never knew you had one heh

Ford, Super,
I think what we are butting heads about here is the different approach to consolidiated vs discrete component implementation. Intel seems to be focusing on a very device centric approach, wherein the the individual specific purpose device needs to have the highest connectivity throughput. AMD seems to be focused on an internalization of the distributive processing, or non-specific approach to the same, thus requiring more "local" resources, at lower speeds, while allowing for better diversity in how the cores inter-relate.

I'm not a fanboy of either, but I can recognize the value of both approaches.
Right now the Intel, with huge L3 caches cost about 5x what a AMD with none run.

Gerritt

Good to see you too Gerritt

Well said. I still think Ford had some good valid points in this manner. But I think you hit the nail on the head. You make a lot of sense. Great post Gerritt

Gerritt said:
Um, the only Intel processors with an L3 (off-die) are the likes of Itanium 2. AMD Phenom/quad-core Opterons have an L3 cache. I assume you meant L2...it boils down to AMD's exclusive 65nm fab versus Intel's inclusive 65nm fab. When you take in consideration the transistor count of the Phenom 9850 (approximately 450m[1]) compared to the Core 2 Extreme QX9770 (more than 800m[2], approximated 900m) and the fabs, Intel's spending about 38% greater cost to manufacture. That is reflected in Intel's price and performance. Yes, Intel's processors are more expensive. Yes, you get more performance because of it and because of the fact Intel is a household name while AMD still is not, they could get away with that 38% even when performance is about the same (e.g. Athlon 64 vs. Pentium 4).

AMD is doing a good job at keeping the transistor count down but at the same time, a high transistor count (ehm, ridiculous amount of cache) really would do more harm than good with AMD's memory scheme. Perhaps, AMD's exclusive memory caching has been outlived and it's time for them to move towards massive, inclusive memory banks like Intel. Perhaps not. It remains to be seen. If AMD continues a downward trend while Intel climbs higher, it may very well be the ghost in the machine.

1. http://www.gearlive.com/news/article/q208-amd-announces-new-ph...k-edition/
2. http://www.overclockersclub.com/reviews/intel_qx9770/2.htm

AMD stated they are moving far away from the Athon 64/Phenom architecture and creating/designing something completely different and new. I'm looking for the link where I got this info from.

Super XP said:
They better!

Yup, they have no choice. I hope Intel is working on the successor to x86-64 too. IA-64 is the right idea but it needs to be modified to be more consumer friendly. Basically, make a thoroughbred 64-bit CPU and tack a x86-64 translator on to it. That way, old software can be used and a new instruction set carries us into the distant future. I hope to see something like that by 2020. x86-64 simply has too much baggage on it.

FordGT90Concept said:

i'm sure it will happen eventually... but can you imagine the enormous fiasco when they finally make the switch and legacy software is no longer supported because it is all coded for an entirely different instruction set?

Well just look at the USB 3.0 fiasco Intel is causing. They are not playing a fair game in this industry. Bad enough they screwed up USB 1.0 back in the day.

Jim H said:
I know, x86 won't go out quietly but it needs to happen at some point. That's why I suggested a translator so that 16-bit, 32-bit, 64-bit (x86-64), and new 64-bit (new instruction set) could all be fully supported. When 128-bit is on the horizon, it could be extended to support it. When it's time to make the move to new 128-bit instruction set, the legacy x86-64 chipset could be scrapped and the 64-bit instruction set be used. This is at least 10 years off and by then, x86-64 should basically disappear or be segregated to old hardware. We need as much legacy support as possible but we also need to move ahead.


Super XP said:
What fiasco? Intel made almost every standard in popular use today from PCI Express to ATX to USB. Sure, they don't always hit the nail on the head (e.g. BTX) but they are an undeniable force in the industry.

Intel plays games with USB3.0
http://www.theinquirer.net/gb/inquirer/news/2008/05/09/intel-p...games-usb3

Super XP said:
Here is that LINK:
AMDs eight-core is a native one
http://www.fudzilla.com/index.php?option=com_content&task=...p;Itemid=1

You all know that I've been advocating the use of multiple back-end RISC processors with a OS/SW friendly CISC front-end for years now.
Heck if Cray, DEC, Silicon Graphics, and even software driven dirivatives (like DNA Computing, or Beowolf) have been around for SO many years, it's about time that this type of technology starts getting implemented on the desktop. Instead of having GL4, DX10, etc., implemented in SW, have it in firmware that directly accesses the back end cores. Faster, more flexable, and a easier upgrade all around.

Gerritt

Doesn't that create potentially major security holes though? Video cards can access the system memory and as such, could provide as a back door for pretty much everything good and bad.